#GRASS IRC Log - 2008-08-01

For logs after Feb 3, 2007, all times are GMT-8. Prior logs are GMT-9.
Back to Logs
00:44:19 CIA-11: neteler * r32441 /grass-addons/contributors.csv: +William Richard
00:44:20 sigq: grassfeed: Changeset [32441]: +William Richard <https://trac.osgeo.org/grass/changeset/32441>
02:08:44 sigq: grassfeed: Ticket #241 ( defect reopened ): v.proj: broken ( ? ) -z description and missing -l description in wxGUI <https://trac.osgeo.org/grass/ticket/241#comment:2>
03:31:57 CIA-11: martinl * r32442 /grass/trunk/vector/v.proj/main.c: v.proj: switch -l flag label/description, trac #241
03:33:09 sigq: grassfeed: Changeset [32442]: v.proj: switch -l flag label/description, trac #241 <https://trac.osgeo.org/grass/changeset/32442>
03:42:55 CIA-11: martinl * r32443 /grass/trunk/gui/wxpython/gui_modules/menuform.py: wxGUI: fix flags label/description in dialog
03:42:56 sigq: grassfeed: Changeset [32443]: wxGUI: fix flags label/description in dialog <https://trac.osgeo.org/grass/changeset/32443>
03:45:23 CIA-11: martinl * r32444 /grass/branches/develbranch_6/gui/wxpython/gui_modules/menuform.py: wxGUI: fix flags label/description in dialog ( merge from trunk r32443 )
03:46:03 CIA-11: martinl * r32445 /grass/branches/develbranch_6/vector/v.proj/main.c: v.proj: switch -l flag label/description, trac #241 ( merge from trunk r32442 )
05:34:34 CIA-11: martinl * r32446 /grass/trunk/vector/v.in.dxf/ ( 12 files ): v.in.dxf: message cleaning, -l flag: print also GRASS compliant name ( used by wxGUI )
05:39:37 CIA-11: martinl * r32447 /grass/trunk/vector/v.in.dxf/ ( layer_list.c main.c ): v.in.dxf: cosmetics in layer list format, update r32446
05:39:45 sigq: grassfeed: Changeset [32447]: v.in.dxf: cosmetics in layer list format, update r32446 <https://trac.osgeo.org/grass/changeset/32447>
07:07:05 CIA-11: martinl * r32448 /grass/trunk/vector/v.in.dxf/main.c: v.in.dxf: don't overwrite output map if not specified by 'output' argument
07:08:01 CIA-11: martinl * r32449 /grass/branches/develbranch_6/vector/v.in.dxf/main.c: v.in.dxf: don't overwrite output map if not specified by 'output' argument ( merge from trunk r32448 )
07:28:18 CIA-11: martinl * r32450 /grass/trunk/gui/wxpython/ ( 5 files in 2 dirs ): wxGUI: initial version of dialog for multiple dxf file import ( each layer is imported as separate vector map )
08:02:01 CIA-11: martinl * r32451 /grass/trunk/gui/wxpython/gui_modules/ ( gdialogs.py goutput.py ): wxGUI: dxf-dialog cosmetics, import only checked layers from the list, optionally add imported layers into layer tree
13:41:51 RMatev: _wolf_: all the problems are because of a tolerance constant
13:42:12 : <_wolf_> say what?
13:42:13 RMatev: if I increase it, it works fine on the lines you gave me yesterday
13:42:20 : <_wolf_> ahah
13:42:46 RMatev: floating point sucks
13:43:00 : <_wolf_> yeah. it is a bit inaccurate :P
13:43:11 RMatev: I'm thinking of using more acurate calculations in certain spots
13:43:14 : <_wolf_> what is this constant?
13:43:17 : <_wolf_> what does it do?
13:44:05 RMatev: it basically says that two points in the same EPSILONxEPSILON square are the same
13:44:37 RMatev: I should probably define the behavior better
13:44:43 RMatev: do some math...
13:44:58 : <_wolf_> : ) it sounds like it would need it...
13:45:10 RMatev: I was thinking of getting the constant from the input line
13:45:20 : <_wolf_> that is also one possibility
13:45:25 : <_wolf_> with a reasonable default
13:45:40 RMatev: yep, but still it sucks
13:45:45 RMatev: that approach
13:45:50 RMatev: cause I tried...
13:46:45 RMatev: Is there a possibility that I can use <big> floating point numbers in my segment intersection function
13:46:48 : <_wolf_> well it puts more responsibility on the user
13:47:14 RMatev: this will certainly solve a part of the problem
13:47:15 : <_wolf_> you mean long doubles 64-bit floating points?
13:48:00 RMatev: I mean floating point numbers with arbitrarily long mantisa
13:48:16 RMatev: there are such libraries there
13:48:50 : <_wolf_> ah you mean adding a dependency... hmm or are these libraries standard to ( most systems )?
13:48:54 RMatev: actually I only need precision which is twice more than double
13:49:22 RMatev: I had an article but I lost it
13:49:29 RMatev: I''l do some research
13:49:43 RMatev: probably they are not
13:50:44 : <_wolf_> long double is I think exactly that
13:51:50 RMatev: http://en.wikipedia.org/wiki/Long_double
13:51:51 sigq: Title: Long double - Wikipedia, the free encyclopedia ( at en.wikipedia.org )
13:51:56 RMatev: not enogh
13:52:15 RMatev: 10 bytes versus 8 for double
13:53:49 RMatev: I'll try to figure something out. We want as little user responsibility as possible
13:53:51 : <_wolf_> so you want 64-bit mathematics
13:53:58 : <_wolf_> yes that is true
13:54:04 RMatev: more like 128 bit
13:54:35 : <_wolf_> hmm well that means pue software implementation
13:54:43 RMatev: yes
13:54:55 : <_wolf_> there should be a gnu scientific library or something like that
13:55:31 : <_wolf_> http://www.gnu.org/software/gsl/manual/html_node/
13:55:32 RMatev: well don't bother to search. I'll think whether its absolutely necessary
13:55:32 sigq: Title: GNU Scientific Library -- Reference Manual ( at www.gnu.org )
13:55:52 : <_wolf_> I think that GRASS might already use that for some BLAS stuff
13:55:55 : <_wolf_> I'll check
13:55:57 : <_wolf_> hang on
13:56:43 RMatev: thanks
13:57:04 : <_wolf_> ah no it uses a frotran BLAS implementation
13:57:51 : <_wolf_> nope but I think that GSL could be a possibility
13:58:45 : <_wolf_> hmm maybe it contains GSL
13:58:55 : <_wolf_> I have a strong feeling of GRASS GSL connectivity
14:01:00 : <_wolf_> Atleast the Native wingrass contains GSL
14:01:03 : <_wolf_> http://grass.itc.it/grass63/binary/mswindows/native/
14:01:04 sigq: Title: Index of /grass63/binary/mswindows/nativeWinGRASS Project Web Page ( at grass.itc.it )
14:02:23 RMatev: aha
14:02:34 RMatev: ok I'll check out that gsl
14:04:44 : <_wolf_> hmm gsl is not used by GRASS
14:08:31 : <_wolf_> I think that you should probably ask on the devel list
14:09:00 : <_wolf_> people there might know more about the problem and maybe know of a solution
14:09:04 RMatev: yep, after I check more precision is absolutely necessary
14:12:24 RMatev: I need to find out what is the highest ratio ( max map feature coordinate )/( map detail ) for all maps
14:13:50 RMatev: map detail is the distance at which two points should become the same
14:14:57 RMatev: actually map detail is the cell size of the grid, to which all map features are snapped
14:16:14 : <_wolf_> but vector calculations are not affected by the cell size
14:17:52 RMatev: I didn't say it right
14:18:43 RMatev: for example see the map you gave me yesterday
14:19:37 RMatev: 197437.69470405 166394.08099688
14:19:37 RMatev: 404361.37071651 166394.08099688
14:20:35 RMatev: all vertices of the lines are given with 0.00000001 acuracy
14:21:16 : <_wolf_> well that I guess is depending on projection and floating point accuracy
14:21:51 : <_wolf_> 1.0 in long-lat is a whole lot more then 1.0 in a transverse-mercator projection
14:22:17 RMatev: yep
14:23:30 RMatev: looking into the lines you gave me is making me thing that using some kind of greater precision is inevitable
14:23:37 RMatev: if we want to do things right
14:23:58 : <_wolf_> hmmh...
14:24:32 : <_wolf_> well if it is unavoidable, then, but let's take this discussion to the mailinglists
14:25:14 RMatev: I'll do that as soon as I know exactly what I want : )
14:25:48 : <_wolf_> : )
14:26:01 RMatev: d = ( ax2-ax1 )*( by1-by2 ) - ( ay2-ay1 )*( bx1-bx2 );
14:26:02 RMatev: d1 = ( bx1-ax1 )*( by1-by2 ) - ( by1-ay1 )*( bx1-bx2 );
14:26:02 RMatev: ra = d1/d;
14:26:11 : <_wolf_> my point was that someone on the list might help you in finding what you want ; )
14:26:37 RMatev: may be, but I want to try myself first
14:26:43 : <_wolf_> :D
14:26:57 RMatev: see the caclulations above
14:27:35 RMatev: all ax1, bx2, ... are doubles. When we multiply we lose precision
14:27:43 RMatev: then again we lose it
14:28:44 : <_wolf_> right, but I don't think there is a way to reduce the number of multiplications ( at least I can't think of any directly )
14:29:16 RMatev: yes
14:29:41 : <_wolf_> so a1 to a2 is one segment and b1 to b2 is another?
14:29:50 RMatev: if there was a trick, the one who wrote segment intersection would have thought it
14:30:00 RMatev: yes thats the case
14:32:41 : <_wolf_> and If I've understood the floating point numbers correctly the only solution would be to increase the number of bits used for each number
14:32:50 : <_wolf_> which in turn will make the calculation slow
14:35:53 RMatev: yes
14:36:10 RMatev: or, there is another option
14:37:05 RMatev: first snap input lines so that their coordinates use only half of the mantissa bits in double
14:37:26 RMatev: then do the calculations with double types
14:37:37 RMatev: but that way we lose information
14:38:32 : <_wolf_> that is not really feasible since in long-lat the numbers can be very small => everything is 0
14:39:04 : <_wolf_> but now I most go to sleep
14:39:13 : <_wolf_> talk tomorrow, ok?
14:39:18 RMatev: ok
14:39:25 RMatev: good night
14:39:33 : <_wolf_> good night
14:57:09 CIA-11: glynn * r32452 /grass/trunk/scripts/d.rast.edit/d.rast.edit.tcl: Set weights on .overview, not .
14:58:46 sigq: grassfeed: Changeset [32452]: Set weights on .overview, not . <https://trac.osgeo.org/grass/changeset/32452>
23:56:03 sigq: grassfeed: Ticket #248 ( defect created ): v.info vs. r.info: different region output formats <https://trac.osgeo.org/grass/ticket/248>